async-oneshot-0.3.2 has been yanked.

async-oneshot

A fast and small async-aware oneshot channel.

Features:

Probably the fastest oneshot channel in the world (see 'Performance')
Tiny code, only one dependency and a blazing quick build.
Complete no_std support (with alloc for Arc).
Unique feature: sender may wait for a receiver to be waiting.

Usage

#[test]
fn success_one_thread() {
    let (s,r) = oneshot::<bool>();
    assert_eq!((), s.send(true).unwrap());
    assert_eq!(Ok(true), future::block_on(r));
}

Performance

Here are benchmark numbers on my Ryzen 9 3900X:

test create                ... bench:          56 ns/iter (+/- 2)
test create_send           ... bench:          55 ns/iter (+/- 5)
test create_send_recv      ... bench:          42 ns/iter (+/- 0)
test create_wait_send_recv ... bench:         109 ns/iter (+/- 3)

Here are the same benchmarks on my 2015 macbook pro:

test create                ... bench:         122 ns/iter (+/- 12)
test create_send           ... bench:         122 ns/iter (+/- 21)
test create_send_recv      ... bench:         126 ns/iter (+/- 8)
test create_wait_send_recv ... bench:         232 ns/iter (+/- 29)

Most of the time is actually taken up by measurement overhead and the last bench is probably benching futures_lite::future::block_on more than it's benching this library. We are, in short, very fast.

Compared to other libraries

The oneshot channel in futures isn't very fast by comparison.

Tokio put up an excellent fight and made us work hard to improve. In general I'd say we're slightly faster overall, but it's incredibly tight.

Note on safety

Yes, this crate uses unsafe. 10 times. Not all of it is performance gaming. Please audit carefully!

async-oneshot 0.3.2

async-oneshot

Usage

Performance

Compared to other libraries

Note on safety

See Also

Copyright and License